TokEnable

FinOps CLI for LLM API costs. Scan code, estimate token spend, recommend cheaper models, and block over-budget PRs.

Quick Start

pip install tokenable
tokenable estimate .

┌──────────────────────────────────────────────────────────────────────────┐
│                          TokEnable Estimate                                │
├─────────────┬──────┬───────────┬──────────────────┬─────────┬────────────┤
│ File        │ Line │ Provider  │ Model            │ $/call  │ $/month    │
├─────────────┼──────┼───────────┼──────────────────┼─────────┼────────────┤
│ src/chat.py │   42 │ anthropic │ claude-opus-4-6  │ $0.0450 │ $45.00     │
│ src/sum.py  │   15 │ openai    │ gpt-4o           │ $0.0035 │  $3.50     │
└─────────────┴──────┴───────────┴──────────────────┴─────────┴────────────┘
Total estimated monthly cost: $48.50

Features

CLI Commands

tokenable estimate .                    # Estimate costs
tokenable check . --max-monthly-cost 500  # CI budget gate
tokenable audit .                       # Find savings opportunities
tokenable fix . --dry-run               # Preview model swaps
tokenable diff . --base main            # Cost diff between branches
tokenable price anthropic claude-sonnet-4-6  # Look up pricing
tokenable calibrate . --provider openai # Calibrate from production
tokenable init                          # Set up config + hooks
tokenable update-pricing                # Refresh pricing data (--dynamic for live)

CI Integration

# .github/workflows/tokenable.yml
- run: pip install tokenable
- run: tokenable check . --max-monthly-cost 500

Fails the build if estimated costs exceed your budget. See CI Integration docs.

Pricing Updates

TokEnable ships with bundled pricing data updated each release. For the latest prices:

tokenable update-pricing              # Refresh from static provider data
tokenable update-pricing --dynamic    # Fetch live prices from LiteLLM community DB

A GitHub Actions workflow runs daily to keep pricing in sync automatically.

MCP for AI Agents

TokEnable runs as an MCP server for Claude Code, Cursor, and other AI agents:

{
  "mcpServers": {
    "tokenable": {
      "command": "python",
      "args": ["-m", "tokenable.mcp"]
    }
  }
}

Tools: suggest_model, estimate_cost, audit, apply_recommendations. See MCP docs.

Configuration

Create tokenable.config.json:

{
  "defaultVolume": 1000,
  "ignore": ["**/test/**"],
  "overrides": [
    { "pattern": "src/batch/**", "volume": 50000 }
  ],
  "budgets": {
    "block": 0.05,
    "maxMonthlyCost": 500
  }
}

Supported Providers

Full list: Supported Models

Documentation

• Installation
• Quick Start
• CLI Reference
• Supported Models
• CI Integration
• MCP Integration
• FAQ
• Contributing

Development

git clone https://github.com/pardhusrepo/tokenable.git
cd tokenable
pip install -e '.[dev]'
make test

License

Apache-2.0